Weblog

A few weeks ago, I gave a talk on something that sounds simple but is actually fairly complex: how to handle long-running MCP tool calls. By default, MCP tool calls are synchronous - client requests, server accepts, client blocks until server responds. If you’re querying a database or making an API request, the client wait time isn’t very noticeable, because the tool call typically completes fast. However, there are some cases where this approach doesn’t work so well. Imagine a client calling a tool to prepare a report. The server needs to search for the data, extract and transform it, analyze it, summarize information and prepare conclusions, generate the final report in a standardized format, and return it to the client. This can take a while. Think about video conversion, audio transcription, browser automation, database migration, multi-step agent workflows - they all have the same problem. ...

I’ll be speaking about the new MCP Tasks extension at the Virtual MCP Community Event for AI Builders by the Agentic AI Foundation. The MCP Tasks Extension (Thu Aug 06 2026 at 21:30 IST) What happens when an agent calls a tool to do something that takes 10 minutes…or an hour? That’s where the new MCP Tasks extension comes in. This talk covers how MCP handles long-running work, what changed in the 2026-07-28 spec, and a live demo. ...

If you wrote an MCP server a few months ago and it’s suddenly looking a little old-fashioned…welcome to the club! The latest MCP specification update (2026-07-28) includes a number of changes: a new extensions framework, OAuth/OpenID authorization updates, a formal deprecation policy, and many more. But perhaps the biggest change is the removal of connection-level sessions and initialization handshakes. Basically, sessions are out, stateless requests are in. Under the previous approach (2025-11-25), an initialization handshake was mandatory, but assigning a session ID and persisting per-client state was an implementation choice, not a requirement of the protocol itself. With the acceptance of SEP-2567 (explicit state handles) and SEP-2575 (stateless MCP), the protocol drops session support entirely. Now, each request is self-contained and context must be passed explicitly in request metadata and tool call parameters. ...

AI can make financial analysis easier, helping to identify patterns, summarise data, and test assumptions - but only if you’re willing to give it your data. I already had all my personal bank statements and portfolio summaries queued up in CSV, ready to go. But there was one problem: there was no way I was uploading all that data to the cloud. So I decided to try a local model instead. My goal was simple: build a workflow where every CSV, every prompt, and every generated report stayed on my own laptop. I knew this would be slow - but I was happy to sacrifice speed for privacy. ...

Yesterday, I woke up to an email inviting me to join the AAIF Ambassador program. Needless to say, I accepted. Agents are having their moment. Everyone is racing to ship them, but there’s still no agreement on how they should communicate, discover capabilities, or share context. We’re building platforms for them before we’ve defined how they coordinate. That’s where the Agentic AI Foundation (AAIF) comes in. Part of the Linux Foundation, AAIF is the neutral, open home for agentic AI projects and standards. Currently, this includes: ...

My talk on sharing context between agent swarms has been accepted for MCP Dev Summit Mumbai 2026. Orchestrating Agent Swarms with MCP, Sandboxing and Shared Filesystems (Sun Jun 14 2026 at 17:50) The moment you go from one agent to three, you hit a set of problems that MCP itself doesn’t solve and that most orchestration frameworks only paper over. This talk walks through those problems using a concrete example of a code review swarm. It demonstrates how agents can collaborate through a shared filesystem rather than by passing context in prompts. ...

I’ll be running a hands-on workshop on integrating coding agents with sandboxes at Open Source Day 2026. Give Your Coding Agent a Perpetual Sandbox for Isolated Execution (Sat Apr 04 2026 at 10:30) This hands-on lab teaches participants how to create isolated compute environments for agents using sandboxes. These sandboxes provide airtight isolation and near-instant boot times, giving agents fast, deterministic environments to run arbitrary code without compromising other parts of the system. ...

My workshop on building modern CI pipelines in Python has been accepted for PyCon Thailand 2025. Build Modern CI Pipelines in Python (Fri Oct 17 2025 at 13:55) This hands-on workshop explores how to build modular, language-agnostic CI pipelines in Python using Dagger, an open-source runtime created by the founders of Docker that treats pipelines as code. It covers building CI pipelines using Python (no more YAML!), reusing logic across teams and projects with Dagger modules, and integrating Go, TypeScript, and other languages into Python pipelines. ...

I’m pleased to announce that my talk on self-healing pipelines has been accepted for AI_dev Europe 2025. The Self-Healing Pipeline: How Dagger + LLMs Are Making Traditional CI/CD Obsolete (Fri Aug 29 2025 at 16:10) This talk discusses an innovative approach to building agentic software for CI/CD: combining Dagger’s programmable open source container engine with LLMs to create self-healing pipelines. It also demonstrates a live example of this approach in action, showing how an LLM can analyze pipeline failures and automatically generate fixes. ...

I’ll be in Hyderabad and speaking at the Cloud Native Hyderabad meetup in two weeks. Introduction to Dagger (Sat Apr 19 2025 at 10:25) This session introduces Dagger, the open source programmable container engine by the founders of Docker. It explains what Dagger is, what it can be used for, and how using it offers significant advantages over traditional CI/CD implementations. The session also includes a hands-on demonstration of live-coding a CI pipeline using Dagger. ...